Ml-Mining: Machine Learning Surrogate Modeling for Environmental Risk Assessment and Water Quality Prediction at Mining Sites

Introduction

Mining waste management represents one of the biggest environmental and socio-economic concerns for mine operators, governments and citizens due to the chemical reactivity of such waste and the potential for releasing toxic drainage enriched in metal(loid)s and acidity with the severe risk of damaging aquatic ecosystems in the vicinity of mining activity. Reactive transport models represent established tools to predict the quality/quantity of such polluted leachates, and to examine future reclamation and remediation scenarios. Yet, intense computational burden, required to solve a multitude of processes and their multilevel coupling, makes it prohibitive to adopt these process-based models with essential level of details in site-scale risk assessment studies. To overcome these limitations, we propose the development of surrogate models, which can predict the output of complex models in timescales that are orders of magnitude faster compared to the corresponding process-based models. The method relies on the application of machine learning, deep learning, and other data science/inverse modeling techniques to estimate the relationship between the inputs and outputs of a computationally expensive model. We will perform detailed investigations to develop surrogate modeling tools tailored for mining systems, to validate the methodology in increasingly complex problems involving different spatial scales, and to test the applicability of such approaches in stochastic risk assessment frameworks. The proposed project will be executed within a duration of 4 years and the applied grant will support a postdoctoral and a doctoral researcher, who will conduct the proposed research activities and tasks. The project outcomes will advance the scientific understanding and provide innovative solutions for practitioners and authorities to improve waste management and environmental risk assessment practices.

Return to projects